Expand benchmarks to include all tests

[L-as]

We can simply change the testing implementation. See examples/Utils.hs.

[TotallyNotChase]

Not all the tests should be (or can be) benchmarked though, right? We should probably include a bench + test utility but also allow raw tests.

[L-as]

I think they all can be benchmarked.

[srid]

Some tests evaluate more than one program. Might require some reorg here (for the foo.bar.qux hierarchical naming to make sense).

[L-as]

Well it's always either one or two. We could call the benches "name:1" and "name:2" when there are two.

[srid]

We can pull in the deterministic benchmarks into tests via golden testing (which also could cover the printTerm stuff, in principle).

It also solves the problem of performance regressions being overlooked (as a result of no one bothering to consistently check the Hercules CI effects output for benchmark diff), because now the CI would fail and the benchmark goldens will have to be reviewed/updated by the PR author. cf. #123

[srid]

This is only partly done; and will be completed as part of #302